Electroencephalography
-
Upload
andresvaldez -
Category
Documents
-
view
214 -
download
0
description
Transcript of Electroencephalography
7/21/2019 Electroencephalography
http://slidepdf.com/reader/full/electroencephalography-56dd8d2fbfe6d 1/21
6Electroencephalography
CONTENTS
6.1 Basic Operational MechanismsPhotic Stimulation
6.2 Signal ProcessingElectrical Coupling of Tissue/Skin and Electronics • Amplifier
• System Calibration • Analog-to-Digital Converter (ADC)
• Digital Signal Processing • Data Display • Amplitude Mapping
• Data Storage • Visual Analysis • Quantitative Analysis
6.3 Detecting MalfunctionsDetecting Focused Malfunction • Epilepsy • Coma Staging
• Brain Death Diagnosis • Intraoperative Brain Function
Monitoring
References
Additional Recommendations, Standards, and Further Reading
6.1 Basic Operational Mechanisms
The electrical potentials generated by the brain’s neural activity can be observed at the scalp using
appropriate amplification techniques. The measured signal is called the electroencephalogram (EEG). It
reflects global brain function, rather than brain function related to the performance of specific cognitive
tasks. Within the framework of everyday neurological examination, therefore, the EEG serves to provide
initial information about global brain condition. For clinical examination purposes, the EEG is recorded
over a period of approximately 15 to 20 min, with the patient sitting relaxed in a comfortable chair,
keeping his or her eyes closed as illustrated in Figures 6.1 and 6.2.
The activity thus measured is called the background EEG, to distinguish it from activity measured
while the patient is engaged in processing of specific external or internal stimuli. Advanced applicationsof EEG recording such as sleep staging, brain function monitoring in the intensive care unit, and so forth
can lead to recording periods of several hours or even days. Additionally, with certain special patient
groups, the brain’s electrical signal may even be recorded from the surface of the cortex or from inside
the brain, but discussion of such specialized recordings is beyond the scope of this chapter.
In order to serve as an index of the function of a spatially distributed neural network, the EEG must
be recorded from multiple measurement positions distributed over the scalp, resulting in a number of
different measured signals. For routine applications, the measurement positions are arranged according
to an international standard called the 10-20 system (Jasper, 1958; see Figure 6.3). This standardized
system facilitates the comparison and interpretation of EEG from different recording sessions and/or
patients. The 10-20 system comprises 19 scalp electrodes and 2 ear electrodes. Accordingly, modern EEGdevices record at least 21 different EEG signals, referred to as channels. For advanced applications, systems
with 32 channels or even more (up to 512) are on the market. The American Electroencephalographic
Hermann HinrichsUniversity of Magdeburg
© 2004 by CRC Press LLC
7/21/2019 Electroencephalography
http://slidepdf.com/reader/full/electroencephalography-56dd8d2fbfe6d 2/21
Society (1994) has published guidelines for the nomenclature of electrode locations that goes beyond
the 10-20 system of Jasper.
Figure 6.4 shows some typical normal background EEG waveforms. The fluctuations in the waveforms
typically cover an amplitude range of 50 µV, peak to peak. In certain cases, however, voltage fluctations may exceed 200 µV. While the EEG contains random noise, it also contains systematic frequency components,
with the main component in most cases being a frequency of about 10 Hz (the so-called Alpha rhythm). For
clinical applications, EEG frequency components ranging from 0.5 to 30 Hz are typically of most interest.
To achieve proper signal representation and interpretation, therefore, a recording system bandwidth of 0.5
to about 70 Hz is required. The recording of the tiny scalp voltages requires both a low-noise large-gain
amplifier and a stable, low-resistance electrical contact between the amplifier and the scalp tissue. The latter
can be attained using appropriate electrodes either glued or mechanically fixed on the skin. With modern
components and electronics, the necessary amplifier specifications can be reached even with restrictive safety
limits protecting the patient from potential risks arising from failure of the electronics or the system operator.
The voltages observed at the electrodes are actually a combination of the EEG and potentials induced by the ambient environment. Among these external potentials are the 50- or 60-Hz line (mains) frequency and
electrostatic charging of the patient’s body. In addition, static electrode potentials also contribute to the total
voltage picked up at the amplifier input. The amplitude of these external potentials usually exceeds the EEG
by several magnitudes. Therefore, differential amplifiers are employed that magnify only the differences
between EEG signals picked up at pairs of electrodes, and not the absolute amplitudes (see Figure 6.5).
FIGURE 6.1 Recording an EEG.
FIGURE 6.2 Basic components of an EEG device.
© 2004 by CRC Press LLC
7/21/2019 Electroencephalography
http://slidepdf.com/reader/full/electroencephalography-56dd8d2fbfe6d 3/21
The external potentials therefore tend to cancel each other out because they are almost identical at all
electrode sites. There are no electrically neutral points on the surface of the head, or anywhere else on
the body, so there is no physiologically canonical way of combining the various electrodes into pairs.
Instead, the EEG has been examined using a number of different electrode montages, meaning that the
EEG can be recorded with different types of electrode pairings. Each type of pairing yields different
measured EEG signals, and thus a different view of EEG topography.Although the set of montages used within a particular clinical EEG laboratory is typically standardized,
there are still no standard montages with international acceptance across laboratories. Traditionally there
are two groups of montages that are called bipolar and unipolar With bipolar montages, arbitrary pairs
of electrodes are formed for the difference amplification process. With unipolar montages, all electrodes
in a particular hemisphere are paired with one common electrode (which is usually located at the mastoid
or at the ear lap) per hemisphere. The referential scheme is a variant of the unipolar montage in which
all electrodes, regardless of hemisphere, share one common reference electrode. During recording, the
reference electrode is often placed on the mastoid or earlobe. Given the use of modern digital EEG
devices, however, the recorded voltage differences may be recalculated after the recording with respect
to a virtual reference electrode, derived by averaging across all electrodes (a so-called common averagereference). EEG recordings based on a referential montage (often called reference recordings) are preferable
for this reason, because any other montage can be derived from the signals by mere recalculation.
Regarding the electrical contact between the skin and the amplifier input, the impedance should be
(1) negligible with respect to the amplifier input resistance, and (2) almost constant over the relevant
frequency range, especially toward low frequencies that approach direct current (DC). To meet these
requirements, electrically conducting electrodes (usually made of metal) are fixed on the skin by various
techniques. An electrolyte typically consisting of paste or gel provides a low resistance between the
electrode surface and the skin. As an inevitable side-effect, this electrochemical system acts like an
electrical battery generating a constant voltage of its own, the size of which depends on the electrode
material. The difference in this constant voltage between different electrode types can be in the rangeof volts. Therefore, only identical electrodes must be used for EEG recording in order to prevent the
differential amplifiers from being overloaded when fed with voltages that vary widely between pairs
of electrodes.
Further external signal components are often superimposed on the EEG, corrupting the measured
signal. There are both biological and technical sources of these so-called artifacts. Some examples are
FIGURE 6.3 Schematic diagram of the international 10-20 system (Jasper, 1958) for electrode placement and
labeling. The term “10-20” refers to the relative distances between adjacent electrodes as indicated in the figure.
© 2004 by CRC Press LLC
7/21/2019 Electroencephalography
http://slidepdf.com/reader/full/electroencephalography-56dd8d2fbfe6d 4/21
FIGURE 6.4 Example of a normal human multichannel EEG. (Upper) Display according to a unipolar recording,
i.e., all signals refer to the electrical potential at either the left or right ear electrode (A1 and A2, respectively). (Lower)
Same EEG-epoch after re-referencing according to a bipolar montage (according to the electrode labels seen at the
left margin).
FIGURE 6.5 Principle of differential amplifier as used in EEG recorders to remove superimposed common high
voltage activity form pairs of EEG signals.
© 2004 by CRC Press LLC
7/21/2019 Electroencephalography
http://slidepdf.com/reader/full/electroencephalography-56dd8d2fbfe6d 5/21
shown in Figure 6.6. Among the biological artifacts are residual electrocardiogram (ECG), electromyo-
gram (EMG) generated by muscles in the vicinity of the electrodes, electro-oculogram (EOG) resulting
from eye movement, and slowly fluctuating electrode voltages induced by transpiration, which influences
the electrochemical conditions of the electrodes. Technical artifacts have a variety of causes, including
inductive nonsymmetric coupling of the 50/60 Hz line (mains) frequency, which can occur when electrode
impedances are too large, movement of the electrodes causing fluctuating electrode voltages, distortions
of the ambient electrostatic field due to movement of the patient or the technician, unstable electrical
contact between electrode and amplifier input, and bad electrode material due, for instance, to insufficient
cleaning of the electrodes. Despite using differential amplifiers, these artifacts often significantly exceedthe amplitude of the ongoing EEG, and can sometimes completely hide it. The 50/60-Hz artifacts and
the EMG can be particularly problematic in this respect. EEG recording systems are usually designed to
record moderately distorted EEG without clipping, although the signal display may still exhibit clipping,
depending on the graphical scaling used to display the recorded voltages.
If artifacts cannot be avoided despite taking measures such as electrode replacement, the interpretability
of the EEG can sometimes be improved if low- or high-frequency components of the recorded waveform
are filtered out, thereby reducing the impact of the artifact. However, unusual low- or high-frequency EEG
activity may be indicative of pathology, and so it may be advantageous instead to extend the frequency band
compared to the routine setting. Whichever the case, the bandwidth of EEG recorders can be interactively
adjusted by modifying both the lower cut-off frequency f l (traditionally specified in terms of the corre-
sponding time constant , defined as 1/2f l ) and the upper cut-off frequency. Also, if line (mains) frequency
artifacts cannot be removed by improving the electrode impedance, an appropriate notch filter helps to
selectively suppress this component. From a technical point of view, adjustment of these filter specifications
may be done by switching the corresponding amplifier settings or by modifying parameters of the internal
digital post-processing algorithms. With modern EEG systems, digital solutions are usually preferred.
FIGURE 6.6 Examples of different types of artifacts (shown only selected channels per artifact): line (mains) artifact,
i.e., 50 Hz activity; electromyogram (EMG) acitivity generated by scalp muscles if the subject isn’t sufficiently relaxed;
movement of electrodes on the skin; eye movements: the moving eye — as an electrical dipole — generates slowly
varying electrical potentials which are picked up by the EEG electrodes; electrocardiographic (ECG) activity embed-
ded in the ongoing EEG.
© 2004 by CRC Press LLC
7/21/2019 Electroencephalography
http://slidepdf.com/reader/full/electroencephalography-56dd8d2fbfe6d 6/21
Traditionally, EEG signals were written continuously in real-time onto continuous z-folded paper
strips. Since the advent of digital EEG systems, the signals are displayed on high-resolution graphic
displays, both during recording and afterward for detailed evaluation. Hard copies are only written for
selected representative EEG segments as part of the physician’s report. The full dataset is stored digitally
on modern mass storage devices, which serve as stable long-term EEG archives.
For routine applications the EEG evaluation is still based on a visual inspection of the signal traces,
taking into account the spatial and temporal distribution of both the spectral structure and the amplitudes
of the multichannel signal. The resulting report is usually a mixture of quantitative measures (for instance,
the topographical area amplitude and frequency of dominant EEG activity, or the relative distribution
of EEG activity over several frequency bands) and qualitative statements (for instance, the occurrence of
specific signal patterns, ratings of the signals with respect to possible pathology, and comparisons to
previous recordings). There are now many post-processing tools available (for instance, for spectral power
analysis, pattern detection, and correlation analysis), but with respect to routine applications none of
these tools has been widely accepted so far. Nevertheless, these tools have extended the applicability of
EEG recordings well beyond their original application in neurological diagnosis. Among the new appli-
cations for EEG are the analysis of the action of cerebrally active drugs, and the long-term monitoringof cerebral function (see below).
Regarding the technical structure of modern EEG systems, different concepts are on the market. The
system may consist of either (1) a front-end signal-preamplifier and main amplifier sending the acquired
data via a dedicated cable to a dedicated computer providing the further functionality needed for data
storage and evaluation (stand-alone system); or (2) the same system as just described, but with the
dedicated computer connected to a central dataserver via a regular network, and with the server also
being used as a reader station (i.e., for data display, or to allow further computers to access the data for
evaluation); or (3) a front-end system sending the data via network connection directly to a server
displaying and storing the data and/or distributing them to other systems linked via the same network.
Moreover, EEG devices are increasingly being integrated into larger IT networks combining variousbiomedical techniques and at the same time organizing patient data management. Finally, there is a
growing tendency to extend the functionality of EEG systems to include other neurophysiological methods
like evoked potentials (EP), electromyography (EMG), or neurography.
6.1.1 Photic Stimulation
One of the main applications of EEG is the diagnosis of epilepsy. Specific spike-like EEG signals, often
accompanied by a subsequent slow wave forming a so-called spike wave complex , indicate a risk of epileptic
seizure. Therefore, these patterns carry a high level of diagnostic information. However, in many patients
the spikes occur only rarely requiring an unacceptably long recording period to pick up at least a few spikes.It is known that some of these patients are sensitive with respect to repeated flashlight exposure. Under
these circumstances a seizure may be provoked thereby improving the chances of picking up a large number
of epileptic EEG patterns. Therefore, EEG recorders are usually equipped with a flashlight generator, with
a flash frequency tunable over a range of approximately 0.1 to 30 Hz. This generator may be triggered either
by the EEG recorder or by an internal clock. The flash onset is recorded and displayed simultaneously with
the EEG so as to be able to correlate it with EEG variations during evaluation. Besides provoking epileptic
EEG activity, the photic stimulation, if presented at an appropriate frequency, sometimes provokes synchro-
nized EEG activity at the same frequency. This phenomenon is known as photic driving .
6.2 Signal ProcessingIn this section the main elements of modern digital EEG recorders will be described in more detail. Older
analog devices that recorded the signals directly on endless paper strips will not be discussed. An excellent
overview of the traditional analog techniques was provided by Cooper et al. (1981). As shown in Figure
6.7, EEG recording may be subdivided into the following logical stages:
© 2004 by CRC Press LLC
7/21/2019 Electroencephalography
http://slidepdf.com/reader/full/electroencephalography-56dd8d2fbfe6d 7/21
• Electrodes
• Analog preamplifier; safety considerations
• Analog-to-digital conversion
• Digital signal processing
• Signal display on screen and on paper• Long-term data storage and retrieval
The remainder of this section is arranged according to this list of topics.
6.2.1 Electrical Coupling of Tissue/Skin and Electronics
Measuring the EEG requires the setting up of a closed loop passing current from its neuronal source
through the various intervening layers of tissue to the electronic amplifier and back again through
the head tissue to the neuronal source. Within this circuit, the coupling between skin surface and
amplifier input plays a critical role; in contrast to the brain tissue, the skin itself usually exhibits a
high resistance and therefore electrodes must be used in combination with a wet electrolyte in orderto bridge this junction.
Standard electrodes consist of a metal plate covering an area of a few square millimeters. The electrolyte
(saline solution) is either contained in liquid form in a pad wrapping the electrode or it is bound in a
gel or paste filling the gap between electrode and skin. Various methods of fixing the electrodes on the
head surface are available: a grid of rubber strips, special caps with the electrodes being integrated into
the cap textile, gluing using a collodium-acetone solution, paste (i.e., the paste providing the electrolyte
also serves as a fixation aide).
Among the metals used for electrodes are platinum, silver, gold, tin, and refined steel. Although all
these materials are good conductors, they become polarized for biochemical reasons if a constant or only
slowly varying voltage is present. This means that an inverse voltage is generated that subtracts from the
real signal resulting in a decrease of the observed voltage. These electrodes are less suitable for the
transmission of low frequency EEG components (i.e., they act like high-pass filters). For this reason,
silver electrodes coated with a thin silver chloride layer (Ag/AgCl electrodes) are the current standard for
routine applications. The chloride layer fully prevents polarization. A special variant is made by sintering
a mixed powder of Ag/AgCl onto the electrode rather than applying it as a discrete metal coating. This
FIGURE 6.7 Technical block diagram of an EEG recorder. Rather than spending separate analog-to-digital converters
(ADC) for each channel, some manufacturers provide just one ADC scanning all channels periodically by means of
an analog multiplexer located between the ADC and the low pass filters. During recording the incoming actual EEG
is displayed on the screen. For later evaluation, arbitrary multiple sets of EEG signals can be fetched from the archive
and displayed in separate windows.
© 2004 by CRC Press LLC
7/21/2019 Electroencephalography
http://slidepdf.com/reader/full/electroencephalography-56dd8d2fbfe6d 8/21
electrode requires less maintenance because, with the conventional Ag/AgCl electrode, the AgCl coating
is easily damaged during routine use. If so, polarization occurs and the electrode needs to be re-coated
with a new AgCl layer. Some electrode types are shown in Figure 6.8.
Special electrodes are available for particular critical applications: for example, for EEG recordings
during operations, or in the intensive care unit, if patients are unconscious, it is important to apply the
electrodes quickly and stably. For these purposes needles of refined steel are available that are prickedinto the skin. In this context, the drawback of polarization is offset by ease of handling and stability of
contact. For recording EEG in patients with certain types of epilepsy, sphenoidal electrodes consisting of
either a steel, platinum, or Ag/AgCl wire are available. These are inserted through the muscle as close as
possible to the anterior part of the brain’s temporal lobe, permitting the recording of epileptogenic activity
with greater sensitivity. Additional invasive electrode locations are often used for epilepsy diagnosis, such
as the nasopharyngeal and foramen ovale electrode locations. In the context of presurgical epilepsy
diagnosis, the recording of electrical activity at the cortex or even in deeper brain regions is sometimes
required for precise localization of the epileptic focus. In these cases, recordings are made from electrodes
arranged on a grid or strip, as well as multiple electrodes assembled on a single piece of wire, usually
using platinum as the electrode material. Recently, graphite scalp electrodes have been introduced thathave special advantages for EEG recordings made during magnetic resonance tomography (MRI).
As previously mentioned, another electrochemical effect, in addition to polarization, results from the
fact that the unit composed of the electrolyte and electrode is — physically speaking — an electrochemical
element like a battery, generating a DC voltage. This DC voltage is uniquely defined by the kind of
electrode metal, and is several orders of magnitude larger than the EEG voltage. Given identical electrode
FIGURE 6.8 EEG recording electrodes. The electrodes shown in the upper row have to be applied individually using
a dedicated flexible rubber frame (upper right). In contrast, electrocaps are available (lower) where all the electrodes
are automatically placed at their right position once the cap is applied.
© 2004 by CRC Press LLC
7/21/2019 Electroencephalography
http://slidepdf.com/reader/full/electroencephalography-56dd8d2fbfe6d 9/21
types for all recording sites, this DC voltage cancels out as long as differential amplifiers are used.
Cancellation of this voltage is a prerequisite for being able to amplify the EEG by a factor of approximately
10,000 without blocking the amplifiers. However, if different electrode metals are used simultaneously
(leading to different electrode potentials), the difference between the electrode potentials will not be
negligible and may provoke amplifier overload, or drive the amplifier into amplitude ranges in which
linearity is not guaranteed. Mixing of electrode types should therefore be avoided. Alternatively, if
different electrode types cannot be avoided, the total gain may be reduced, at the cost of a reduction in
signal-to-noise ratio.
For a good quality recording, the impedance of the electrode-skin junction should be kept below 5
k W. However, this limit is sometimes too strict for applications in which patients cannot tolerate a
long preparation phase. As a rule of thumb, keeping the impedance below 20 k W still results in
acceptable recording quality. The importance of keeping impedances as low as possible derives from
the following facts:
1. The amplitude of the EEG as measured at the amplifier’s input depends on the ratio of the electrode
impedance and the amplifier input impedance. Osselton (1965) has shown that the relative dropof the measured EEG amplitude is (R1 + R2)/(Rin + R1 + R2), where R1 and R2 are the impedance
of the two electrodes whose voltage difference is being amplified, and Rin is the amplifier’s input
impedance. However, with modern EEG recorders the input impedance is usually much better
than 10 MW resulting in a signal loss below 1% even for electrode impedances up to 50 k W.
2. By the general rules of physics, a resistor R generates a thermal noise voltage Vn which is specified
by the equation
over the spectral bandwidth B, where k is the Boltzman constant and T the absolute temperature.With large resistances, this additional noise contributes significantly to the total measurement
noise, thus degrading the signal-to-noise ratio.
3. Due to electromagnetic and electrostatic coupling, the line (mains) frequency (50 or 60 Hz) can
be coupled into the measurement setup. When the electrode impedances are large, the resulting
currents generate a 50/60 Hz AC voltage at these impedances, thus causing an artifact superim-
posed on the EEG signal. Due to the fact that the differential amplifiers (see above) record the
difference between two voltages picked up at two electrodes, this effect is most pronounced if the
impedances of the two electrodes used for the measurement differ greatly; it does not depend so
much on the absolute values of the electrode impedances.
The ability to measure electrode impedances is usually included as a special function in modern EEG
recorders. For this purpose, a constant alternating current (AC; usually 10 Hz) is fed from the recording
electronics to the electrodes. According to Ohm’s law, the voltage observed under these conditions directly
reflects the resistance. Depending on the various recorder types and suppliers, graphical or alphanumer-
ical methods are used to display the actual resistance values for all electrodes at a glance. For practical
reasons, it is advantageous to have this display at the preamplifier located close to the patient’s head,
rather than only at the main recording unit.
6.2.2 Amplifier
Standard low noise operational amplifiers with high input impedance (>10 MW) are used to amplify thevoltage differences between pairs of electrodes. The amplifier is split into two modules, with the first
stage providing a modest gain of about 10. Before entering the second stage the signals pass a coupling
capacitor that removes potential residual high voltage DC potentials that might occur if electrode poten-
tials are not equal over the electrodes involved (which in practice usually cannot be avoided). The overall
gain in most EEG systems is on the order of 10,000 to 20,000, yielding an EEG amplitude of about 1 volt
Vn kTBReff = 4
© 2004 by CRC Press LLC
7/21/2019 Electroencephalography
http://slidepdf.com/reader/full/electroencephalography-56dd8d2fbfe6d 10/21
at the amplifier’s output. Due to the DC-blocking capacitor between the two modules, the amplifier has
a high pass characteristic with a low frequency cutoff that is defined by the capacitor. Traditionally, the
time constant of this circuit is specified, rather than the lower cut-off frequency. Usual values are 0.03,
0.1, 0.3 (standard), 1, and 3 sec, corresponding to 5, 1.6, 0.5, 0.16, and 0.05 Hz. Short-time constants
facilitate the interpretation of EEG signals when there are large superimposed low frequency components,
due either to artifacts or pathological activity. However, if pathological activity at low frequencies needs
to be evaluated with high sensitivity, for instance, for brain death diagnosis, larger time constants are
required. EEG recorders therefore allow switching between different settings. On the amplifier side, this
is accomplished by hardware switches that select between various capacitor values. An alternative (in
many available systems, additional) approach is to change the effective time constants by modifying the
coefficients of the digital filters during signal post-processing (see below).
Besides providing adequate signal transmission, the amplifiers must be designed to match the safety
demands specified by the IEC 601 standard as formulated by the International Electrotechnical Com-
mission, Geneva, Switzerland (1994). The main goal is to rule out the possibility that the current
flowing from the amplifier input through the tissue exceeds 100 mV, even in the case of a failure of
the electronics. Such protection can be achieved using appropriate resistors between the electrode cableand amplifier’s input pins. In addition, modern EEG amplifiers usually provide full electrical insulation
(using optical transmitters) of the front-end amplifier from subsequent electronics in order to prevent
high voltages entering into the front end. Such decoupling is known as floating input , because there
is no stable relation between the absolute signal amplitude within the preamplifier and the ground
potential of the subsequent stages.
For special applications DC recording units are available that transmit the input difference signals
without any frequency limitations, that is, without any DC-suppressing coupling capacitor. In order
to avoid excessively large amplitudes that would exceed the amplifier’s dynamic range, it is necessary
to provide an individual DC voltage for each channel that is subtracted from the difference signal,
thus compensating for the residual DC component attributable to fluctuating electrode potentials.From time to time these recorders need to be reset interactively in order to adapt the voltage of this
DC compensation signal. Alternatively, a slow voltage follower may track the fluctuating DC,
adapting the compensation signal automatically. However, in a strict sense, this is no longer a pure
DC recorder.
Modern EEG systems are designed as referential recorders, meaning that all electrodes are measured
with respect to one common reference electrode placed somewhere on the head. Accordingly, this
electrode is internally connected to the inverting input pins of all the difference amplifier channels.
One important characteristic of these difference amplifiers is their common mode rejection (CMR)
characteristics, which are defined by the ratio g1/g2, where g1 is the gain applying to the difference
voltage at the two input pins, and g2 is the gain applying to the common voltage at both input pins.A large CMR is a prerequisite for efficient suppression of noise components present at both input
pins. Modern EEG amplifiers achieve a CMR of 80 dB or even better, thereby reducing common noise
components by at least 1:10.000.
The intrinsic noise level of modern EEG amplifiers is about 0.5 mVeff at a bandwidth of 100 Hz
before amplification Adding the noise originating at the electrodes, a total noise floor up to 0.7 mVeff ,
corresponding to an approximate peak-to-peak level of 2 to 3 mV (Gaussian amplitude distribution),
is realistic.
6.2.3 System Calibration
Although modern operational amplifier electronics are very stable with respect to their specifications, it
is still good practice to calibrate the total system each time an EEG is recorded. For this purpose a highly
stable reference signal is provided that can be internally connected to all amplifier inputs on demand.
Using a 1-Hz square wave signal (which is offered by most systems), the main features of the amplifier
and all post-processing steps including filters can be checked at a glance.
© 2004 by CRC Press LLC
7/21/2019 Electroencephalography
http://slidepdf.com/reader/full/electroencephalography-56dd8d2fbfe6d 11/21
6.2.4 Analog-to-Digital Converter (ADC)
Conversion from analog-to-digital signal representation (see Figure 6.9A) requires constraints to be placed
on (1) the spectral bandwidth, (2) the amplitude resolution, and (3) the amplitude range. The sampling
rate and the number of bits/sample (sometimes called the resolution) then needs to be adapted appropriately.
According to the Shannon Nyquist theorem, the minimum sampling rate fs for adequate digital representation
of analog signals is 2*fn, where fn is the highest frequency occurring in the signal. If this rule is violated
(i.e., fs< 2*fn), a component at frequency f > fn will result in a spurious frequency component of a lower
frequency after digitization, with the frequency of the spurious component being given by fn-(f-fn). This
effect is known as aliasing (see Figure 6.9B). To prevent such distortion, analog low-pass filters (anti-aliasing
filters) with an appropriate cut-off frequency are used to suppress high frequencies before digitization. Due
to the limited steepness of the roll-off characteristic of the filter, the sampling rate should be slightly higher
than twice the cut-off frequency fc. Typical values for fc and fs are 100 and 256 Hz, respectively. This
bandwidth is larger than would be necessary for the EEG alone. However, high frequency artifacts such as
electromyographic (EMG) activity sometimes can only be distinguished from EEG with the use of this
larger bandwidth. Some EEG systems extend the analog bandwidth and the sampling rate even further and
go down to the lower rate only after digital filtering and subsequent down sampling (i.e., sampling rate
reduction, also known as decimation) as illustrated in Figure 6.10. Once the signals have been digitized with
a sampling rate of fs, all further digital processing occurs within the limited frequency range of 0 to fn Hz.
FIGURE 6.9 (A) Principle of analog-to-digital conversion. The continuous analog signal is converted into a sequence
of numbers (indicated on the y-axis) representing the signal amplitudes at discrete points in time with a limited
number of different amplitude values. The resulting temporal and amplitude resolution are labeled as Dt and ULSB
in the figure. The sampling rate thus is 1/Dt. (B) Aliasing: A high frequency component (dotted line) mimics a low
frequency component (straight line) if sampled with too low a rate.
FIGURE 6.10 After passing an analog filter of large bandwidth, the signals are digitally sampled with a correspond-
ingly high rate. A subsequent digital filter substantially reduces the bandwidth. Finally, the digital data rate is reduced
accordingly. Usual reduction ratios (the decimation factor ) are 1:4 or 1:16.
© 2004 by CRC Press LLC
7/21/2019 Electroencephalography
http://slidepdf.com/reader/full/electroencephalography-56dd8d2fbfe6d 12/21
Current commercial EEG recorders provide a resolution r of either 12 or 16 bit per sample. The ADC
input range covered by these 4,096 or 65,536 different code words must be determined by a trade-off
between the amplitude range to be coded and the least significant bit (LSB), which determines the
additional quantization noise generated by the ADC. The LSB is the input amplitude step (labeled as ULSB
in Figure 6.9B) corresponding to the numerical difference between two adjacent digital codes, and thus
determines the precision of the ADC. This precision should be kept significantly below the analog noise
level to prevent further degrading of the signal-to-noise ratio. Reasonable LSB values are thus well below
1 mV before amplification. The amplitude range R can the be calculated as LSB*2r-1. For example, assuming
an LSB of 0.5 mV and an r of 12 bit, the range is R = 0.5*4096 mV or ±1.024 mV. This value is sufficient
for ordinary EEG signals, but may still clip large amplitude artifacts. For instance, severe EMG artifacts
occasionally exceed this limit. Also, in the case of DC recordings this bound may be exceeded in unfa-
vorable situations. Two ways of extending the ADC amplitude have been implemented in commercial
systems: either the larger 16-bit resolution is used, extending the range by a factor of 16, or the ADC
input range is arbitrarily shifted toward larger values, at the cost of precision/digitization noise. The
second solution is only acceptable if low amplitude signal components are of little interest.
6.2.5 Digital Signal Processing
Once the EEG has been digitized, all the basic functions of the EEG recorder can be realized by using
numerical algorithms and data transfer operations executed either by the main processor or by dedicated
signal processors. The main functions are: (1) digital filtering and (2) calculation of virtual EEG signals
corresponding to various electrode montages. Filtering is sometimes necessary to facilitate visual EEG
evaluation in cases where abnormal pathological or artifactual signal components at certain frequency
bands overlay (and thereby hide) the background activity under consideration.
6.2.5.1 Digital FilteringWith respect to EEG recording, filtering aims at the suppression of selected frequency ranges, either
toward the upper frequencies (low-pass filtering) or the lower frequencies (high-pass filtering). In addition,
a dedicated filter is usually provided for the selective suppression of one fixed frequency component
(notch filter ) with the goal of rejecting line frequency artifacts (50 or 60 Hz). Ideally, a filter transmits
frequency components within its passband with a gain of 1 (i.e., perfect transmission) and frequencies
within its stopband with a gain of 0 (i.e., perfect suppression). However, for various reasons, real-world
filters do not achieve this perfect performance. Instead, the characteristics of real-world filter algorithms
can be described as follows (see Figure 6.11). Within the passband the gain stays in a range 1 ± e, where
e is known as the passband ripple, whereas throughout the stopband the gain is below d<<1, where d is
known as the stopband ripple. The frequency range between the passband and the stopband is namedtransition band , within which the gain monotonically decreases from 1 to 0. The steepness of this
decreasing characteristic is specified by the roll-off value, usually given in terms of the logarithmic
attenuation (decibel [dB]) per octave. The frequencies defining the edges of the passband and stopband
are also called cut-off frequencies.
Compared to analog filters, digital filters are advantageous for several reasons: (1) they can easily be
designed to work without phase distortions (i.e., they can have a perfect linear phase characteristic); (2)
changing the cut-off frequencies is accomplished by simple modification of the numerical parameters
rather than changing hardware components; and (3) the filter characteristics are strictly identical for all
channels, whereas different analog filters differ with respect to their exact specifications, due to variability
in electronic components. Moreover, digital filters are cheaper because the only investment needed is thedesign, which is easily accomplished using standard tools, and only has to be done once when developing
the EEG recorder. Digital filtering may be applied in real-time, that is, while recording the signals, as
well as later during data analysis and evaluation.
The general algorithm of a linear phase digital filter calculates the filtered signal samples as a weighted
average of a limited number of subsequent samples of the input signal. The type of filter is defined by
© 2004 by CRC Press LLC
7/21/2019 Electroencephalography
http://slidepdf.com/reader/full/electroencephalography-56dd8d2fbfe6d 13/21
the sequence of weighting factors. For instance, if all the factors are all more or less similar, the output
signal will come close to a moving average of the input signal. This filter thus resembles a low-pass filter.
Alternatively, if the successive weighting factors have alternating signs, the filter will behave like a differ-
ence operator (which suppresses slowly varying signal components) and thus act like a high-pass filter.
The formal definition is as follows:
where a j = a p-j , the filter output being designated by y i, the filter coefficients by a j (i.e., the weighting
factors mentioned before), the filter order by p, and input data by x i. The subscript i indicates the
time step (i.e., the recording time between samples is i*Dt , with Dt specifying the sampling interval
1/fs). From a numerical point of view, the various filter types differ only in the filter coefficients a j
and the filter order p. The latter mainly determines the computational effort needed to apply the
filter. Large filter orders are necessary to realize filters with very small passband and stopband ripples,
a narrow transition band and other rigid features. Therefore, the filter characteristic needs to becarefully defined in order to keep the computational demands within reasonable limits. In addition,
with very narrow transition bands (resulting in steep roll-off slopes, which is optimal in terms of
frequency characteristic; see previously), the filter output tends to overshoot (i.e., distortions) in case
of steep signal gradients if the order p is too low. With respect to the EEG, this aspect is especially
important for notch filter design.
FIGURE 6.11Schematic frequency responses of digital low pass (upper figure) and high pass (lower figure) filters.
The Nyquist frequency is indicated by fs. The frequency characteristic repeats periodically toward higher frequen-
cies because the corresponding signals cannot be distinguished from low frequency components due to aliasing
(see above).
y a x i j i j
j
p
= * -
=
-
Â0
1
© 2004 by CRC Press LLC
7/21/2019 Electroencephalography
http://slidepdf.com/reader/full/electroencephalography-56dd8d2fbfe6d 14/21
Although digital filters in principle have completely flexible adjustment of cutoff frequencies, com-mercial EEG recorders usually provide a few preset cut-off frequencies for the passband. Typical examples
are 15, 30, and 70 Hz for the upper cut-off, and 0.053, 0.16, 0.53, and 1.6 Hz for the lower cut-off (the
latter corresponding to time constants of 3, 1, 0.3, and 0.1 sec, respectively).
6.2.5.2 Re-Referencing EEG Signals according to Various Electrode Montages
As previously mentioned, with older analog EEG recording devices, the electrode montage used
during the recordings determined the electrode montage to be used for display and evaluation of the
signals. By contrast, with modern digital EEG recorders, the signals are stored in a form that permits
subsequent reprocessing. Given that the signals have been recorded with a referential electrode
montage to begin with, it is then possible to recalculate the signals to simulate any other possibleelectrode montage.
To illustrate how to calculate an arbitrary EEG difference signal from the raw data, let us consider how
the signals resulting from a bipolar electrode montage can be computed from the signals obtained from
the original referential recording. If we assume that E1i – Ri and E2i – Ri (where i denotes the sequence
of time points 1, 2, etc.) are the digital representations of two referentially recorded EEG channels picked
up at electrodes E1 and E2, with R denoting the reference electrode, the difference signal E1i – E2i is
easily computed as
E 1i – E 2i = (E 1i – Ri) – (E 2i – Ri)
More complicated montages comprising channels from more than just two electrodes can be derived
as well by such simple numerics. Alternatively, virtual references may be generated by combining a group
of electrodes. The popular common average reference is a special example in which the average over all
recorded EEG channels is taken as the reference, thus representing a spatial average EEG. The detailed
definition of this type of reference varies among EEG laboratories.
FIGURE 6.12 Screen shot taken from a recording EEG device.
© 2004 by CRC Press LLC
7/21/2019 Electroencephalography
http://slidepdf.com/reader/full/electroencephalography-56dd8d2fbfe6d 15/21
6.2.6 Data Display
For routine diagnostic applications, EEG evaluation is still based exclusively on visual inspection. In the
past, the EEG traces were continuously written on continuous z-folded paper strips, with a more or less
fixed temporal and amplitude scaling of 3 cm/sec and 70 mV/cm, respectively, 10 sec of traces recorded
per page. Different scalings were available for special applications. With modern systems, the EEG isanalyzed using high resolution computer screens to display the signals. Usually the graphical resolution
is better than 1024 by 768 pixels, thus allowing for a quality that is almost — although still not fully —
comparable to the traditional paper recording. The advantages of this technique over paper recordings
are obvious: signals can be displayed repeatedly at different scalings, color coding permits different signals
to be distinguished, especially if the signals cross each other when there are large amplitudes, the different
channels can be arbitrarily arranged on the screen, and measurement of amplitude and period duration
is supported by providing interactive cursors, and so forth. In Figure 6.12 a snapshot of an EEG recorder
screen is displayed.
Regarding real-time display, the operator can mark several event types (for instance, patient movement,
eye movement, artifact, etc.) by pressing certain buttons or clicking the mouse. The number and meaningof different event codes can usually be defined by the user. Of course, all this marker information is
stored with the raw data so as to be available during later evaluation. During EEG evaluation it is often
helpful to have another EEG at hand (for instance, a previous recording of the same patient) that can
be displayed simultaneously. For this purpose most current EEG systems can open additional graphic
windows and display a separate EEG in each window. This feature may also be used to compare different
sections of the same EEG during evaluation. In addition to the screen display, hard copies may be
generated from selected parts of the EEG in order to serve as traditional printed documents in an EEG
evaluation report. Conventional inkjet or laser printers clearly provide sufficient resolution to print the
EEG traces with high quality.
6.2.7 Amplitude Mapping
The traditional way of displaying the observed EEG is to draw the amplitude of the n channels over time
as n separate traces. This method allows for an excellent temporal analysis but is less suitable for
topographical evaluations. Therefore, a complementary method has been developed showing the ampli-
tudes of all channels at one instant in time, or averaged across a short time interval, for the topographical
region covered by the electrode set (assuming a reference montage). Figure 6.13 shows an example of
such a topographic map. Amplitude values between electrodes are estimated from the measured values
by interpolation techniques, for example, by two-dimensional spline algorithms (see Perrin et al., 1987).
The amplitude values are color or grayscale coded. Alternatively, equipotential lines may be used to
FIGURE 6.13 Display of spatial amplitude distributions applying mapping techniques. Amplitudes between elec-
trodes (marked by dots in the figure) are estimated by means of spatial interpolation algorithms.
© 2004 by CRC Press LLC
7/21/2019 Electroencephalography
http://slidepdf.com/reader/full/electroencephalography-56dd8d2fbfe6d 16/21
display the spatial amplitude distribution. A sequence of maps derived for different time points provides
a concise overview of the spatiotemporal evolution of the EEG.
A crucial issue with amplitude mapping is the selection of the reference electrode. Maps derived using
different reference signals (for instance, the central electrode Cz vs. a common average reference) may
look completely different. Also, artifacts may lead to substantially corrupted maps, with the consequent
risk of a severe misinterpretation if only the map is used for evaluation, rather than using it in conjunction
with the raw underlying signals.
6.2.8 Data Storage
According to official regulations, clinical EEG data must be safely stored over 10 years or longer. The
data volume per EEG session is typically about 10 MB (assuming 21 channels, a 256 Hz sampling rate,
2 bytes per sample, and a 15-min recording time). Despite the constantly growing capacity of magnetic
disk storage devices, this medium is still not large enough to hold all EEG records acquired over several
years. In addition, magnetic disks may “crash” and thus are not acceptable with respect to data safety.
Therefore, current EEG devices are equipped with long-term storage devices. The IFCN recommendationsfor digital EEG recorders (Nuwer et al., 1998) encourage users to use optical disks to archive data, rather
than magnetic tape media. The latter are less stable due to suboptimal mechanical and magnetic char-
acteristics; in particular, there is a risk of erasing data due to ambient magnetic fields in the vicinity of
strong electric currents. Recently, recordable CDs have become popular as storage devices. The stability
of this medium may be sufficient for a 10-year period, but care must be taken to properly handle and
store the CDs because they are not as well mechanically shielded as regular optical disks, which are
permanently housed in individual cassettes.
As a general recommendation, the raw data should be stored on disk without any further processing.
The patient information as well as the all relevant technical parameters, plus event marks, should be kept
within the same file. This guarantees that a particular patient’s EEG file can be retrieved even if thepatient database crashes.
6.2.9 Visual Analysis
In routine clinical EEG applications, the focus of the visual analysis is on (1) the spectral structure of
the EEG, (2) the topographical distribution of the various frequency components, (3) potential hemi-
spheric asymmetries, (4) focused abnormal activity, and (5) spikes or a complex of spikes with a
subsequent slow wave, especially in the context of EEG recorded from epileptic patients. The spectral
evaluation concentrates on estimating the main amplitude and frequency in four basic frequency
bands: Delta (0.5–4 Hz), Theta (4–8 Hz), Alpha (8–13 Hz), and Beta (13–30 Hz). These bands areaccepted with only minor variations worldwide. In most cases the main activity of the adult EEG is
observed in the Alpha band at around 10 Hz. Often one of the main problems is to discriminate
artifactual activity from real EEG. Therefore, in addition to a high quality EEG recording, it is important
to get sufficient information from the technician about potential artifacts that may have been intro-
duced during recording.
6.2.10 Quantitative Analysis
Advanced applications such as monitoring of long-term brain function, sleep staging, evaluating the
effects of drugs on the EEG, and so forth are not feasible by means of visual analysis because thismethod does not involve statistical analysis. Moreover, visual techniques are too time consuming and
prone to error if hours of EEG need to be analyzed routinely. Therefore, a large variety of quantitative
computer-based analysis techniques have been developed. A comprehensive presentation of all relevant
methods exceeds the scope of this paper. Here, though, some methods are sketched that have gained
some practical importance.
© 2004 by CRC Press LLC
7/21/2019 Electroencephalography
http://slidepdf.com/reader/full/electroencephalography-56dd8d2fbfe6d 17/21
6.2.10.1 Spectral Analysis
This technique aims at estimating the spectral distribution of the signal power. Depending on the
application, the result is just one power spectrum per channel, or a sequence of short-term spectra per
channel. For further processing, the spectral data (some 100 numbers per spectrum) are usually concen-
trated in band-power values representing the average power and the median frequency in a small set of frequency bands. Various approaches to estimating the spectra have been proposed. Among these are (1)
calculating and averaging the short-term spectra of consecutive signal segments using the fast Fourier
transform (FFT); (2) parametric spectral analysis by fitting stepwise constant or time-varying linear
stochastic models to the data; and (3) wavelet-based time-varying spectral analysis. A typical EEG power
spectrum derived with method (1) is shown in Figure 6.14.
Spectral analysis is not restricted to the separate analysis of individual signals, but may be used to
analyze the joint spectral power in pairs of signals, resulting in cross spectra. A special normalized variant
of this technique results in coherence spectra which reflect the spectrally resolved correlation between
pairs of signals. Coherence spectra have been used by a variety of authors to analyze functional connec-
tivity between different brain regions.
6.2.10.2 Pattern Recognition
One of the paramount applications of the EEG is supporting the diagnosis of epilepsy. EEG spikes, and/
or spikes combined with a subsequent slow wave, are highly specific in indicating a risk of epileptic
seizures. In many epileptic patients these patterns occur only very rarely or in an unusual morphology,
even under photic stimulation. Hours or even days of continuous EEG recordings are sometimes required
to pick up a sufficient number of spikes for a valid evaluation. Visual inspection is hardly feasible for
these long-term EEG traces. In order to recognize these patterns, automatic procedures have been
developed that detect steep slopes, sudden amplitude increases, unusual local spectral structure, and so
forth. Once these patterns have been identified, a second important step is to discriminate true patterns
from artifacts that frequently mimic these patterns.
FIGURE 6.14 Example of an EEG power spectrum (representing one channel) estimated by averaging over 20
short-term spectra calculated from subsequent 4-sec epochs (one example shown in the upper trace). Note the peak
near 10 Hz representing the dominating 10 Hz background activity seen in most nonpathological EEG recordings.
© 2004 by CRC Press LLC
7/21/2019 Electroencephalography
http://slidepdf.com/reader/full/electroencephalography-56dd8d2fbfe6d 18/21
6.2.10.3 Sleep Staging
EEG recordings are central in sleep analysis, because the main EEG activity shifts more and more toward
lower frequencies with deeper sleep stages. According to rules defined by Rechtschaffen and Kales (1968)
(R&K), the EEG plus some additional physiological signals (electrooculogram [EOG], EMG) are rated
in 30-sec steps. As a result each 30-sec period is assigned to four different sleep stages, plus the rapid eye
movement (REM ) stage. Applying this procedure, the evaluation of an 8-h sleep recording can be com-
pressed into one comprehensive diagram showing the sequence of the 30-sec sleep stages. An example
of this so called hypnogram is shown in Figure 6.15.
Since the introduction of the R&K rules, sleep analysis has relied on mere visual inspection, with each
such analysis requiring several hours of visual inspection. Algorithms have therefore been developed for
automatic analysis. The principal idea is to perform a spectral analysis and classify the low frequency
activity using a discriminant function. In addition, some coherence measures help to discriminate normal
sleep activity from REM activity. Recently, neural network techniques have been successfully applied in
classifying sleep activity. Nevertheless, in many laboratories visual inspection of the signals is still the
gold standard, because many automatic procedures still lack reliability, producing conclusions that differ
from those of the human expert.
6.3 Detecting Malfunctions
The capacity of the EEG to provide a quick overview of global brain function is exploited for routine
neurological examinations, as well as for long-term brain function monitoring during operations and in
the intensive care unit. Here some typical applications are briefly discussed. A more comprehensive
description is found in Niedermeyer and Lopes da Silva (1998).
FIGURE 6.15 Example of a hypnogram (upper trace) resulting from an evaluation of a whole night recording
according to the Rechtschaffen and Kales rules. Besides several channels of EEG, additional signals are included,
among them the electrooculogram (EOG) and an electromyogram (EMG). Each step represents a 30-sec epoch
classified into one of five different sleep stages (plus wakeness). Note the almost periodic repetition of low and deep
sleep stages. In the lower part of the figure some additional results are shown like phases of rapid eye movements
(REM, defining a sleep stage of its own), EMG activity, etc.
0
22:45
1
23:45
2
0:45
3
1:45
4
2:45
5
3:45
6
4:45
7
5:45
8
6:45
9
7:45
10
8:45
Relative
AbsoluteTime
Res
Ekg
Emg 1Myok.
MaMT
4
3
2/RR/2
1/Rem.
Wach
Rem-Act.
(0−8)
Hours from "lights out"
© 2004 by CRC Press LLC
7/21/2019 Electroencephalography
http://slidepdf.com/reader/full/electroencephalography-56dd8d2fbfe6d 19/21
6.3.1 Detecting Focused Malfunction
In a routine neurological examination, the EEG is often used either to exclude or to prove topo-
graphically focused central functional disturbances. For this purpose, the physician looks for topo-
graphically abnormal frequency distributions (especially frequencies below 8 Hz) occurring at a
limited number of electrode sites, thus reflecting a limited brain area. The EEG is not only sensitiveto such disturbances, but can also provide an estimate of severity of malfunction. However, con-
clusions as to the reason for the malfunction (which might be a tumor, an edema, a brain contusion,
etc.) cannot be based on the EEG alone. A classical application for the EEG is estimation of the
severity of head injuries. Also, following head injury the course of central nervous system function
can be noninvasively monitored over hours, days, or weeks in the intensive care unit. Encephalitis
is one of the rare diseases for which the EEG alone can provide strong evidence for its differential
diagnosis, based on specific patterns of focused abnormalities over the temporal lobe that are almost
unique to this disease.
6.3.2 EpilepsyTwo prominent issues in the context of epilepsy diagnoses are (1) to classify the type of epilepsy,
and (2) to identify the epileptic focus driving the pathological activity. Both issues are important
for selecting an appropriate therapeutic approach. The EEG is capable of addressing both questions.
In general, most epileptic patients exhibit specific EEG patterns in a restricted brain area, which
can be observed in a subset of skull electrodes located near these areas. These patterns usually
comprise one or several subsequent spikes (duration < 100 msec), which are sometimes followed
by a slow wave (duration = a few hundred milliseconds). Several variants of this so-called spike wave
complex are specific to particular variants of epilepsy, permitting the use of EEG for differential
diagnosis.
In addition, the topographical distribution of the abnormal activity can help to localize the epileptic
focus. Source analysis techniques may further support this localization. Given a high signal-to-noise
ratio, they are able to relate the surface voltage distribution to the underlying intracerebral neural
sources. One of the main problems here is to observe a sufficient number of interictal spikes within
a routine EEG recording session of only 15 to 20 min. Therefore, long-term EEG recording techniques
are available that use ambulatory devices to pick up these EEG sequences. However, in severe cases in
which only surgical therapy can help, depth recordings may be required to identify the epileptic focus
with sufficient precision.
6.3.3 Coma StagingIn comatose patients, both adequate therapy and prognosis depend on the coma depth and its develop-
ment. The EEG is a sensitive measure for evaluating these. Again, certain morphological and topograph-
ical EEG patterns provide specific evidence of various coma stages.
6.3.4 Brain Death Diagnosis
The spontaneous EEG is one of the most sensitive measures available for brain death diagnosis. According
to official German recommendations, the EEG can prove brain death if (1) there is clinical evidence for
this pathological status, (2) no brain-related electrical activity can be recorded even at an increased
amplifier gain (or larger scale factor for the screen display), and (3) the low cut-off frequency is extendedto 0.16 Hz and the recording time is extended to 30 min. However, artifacts obscuring the underlying
brain electrical activity often preclude a valid EEG interpretation. Also, certain drug states (for instance,
due to barbiturate abuse) as well as a low body temperature may simulate a brain-death EEG, and these
factors must be taken into consideration before EEG recording.
© 2004 by CRC Press LLC
7/21/2019 Electroencephalography
http://slidepdf.com/reader/full/electroencephalography-56dd8d2fbfe6d 20/21
6.3.5 Intraoperative Brain Function Monitoring
Intraoperative EEG monitoring mainly focuses on two fields of application: control of anesthesia and
general brain function monitoring during operations in the brain. Pichlmayr and Lips (1983) and others
have shown that the depth of anesthesia is significantly reflected in specific EEG patterns, which can be
automatically detected and classified in real-time by appropriate computer algorithms. As direct measuresof brain function, in contrast to epiphenomena such as blood pressure and heart rate, these indices allow
for a more sensitive and immediate control of the depth of anesthesia, thereby saving drugs and reducing
side effects. In addition, the risk of unintended wake-ups is minimized.
During surgery on brain vessels (for instance, arteria carotis endarterectomy), a temporary clamping
of the vessel is sometimes required. In such situations, the EEG allows one to check whether the missing
oxygen supply is tolerated by the brain. Potentially impaired brain function can be detected immedi-
ately from pathological EEG signs like decreased amplitudes and/or slowing in the relevant topograph-
ical area. Monitoring of the EEG can be done visually. However, for longer-term monitoring, support
from automatic computer-based procedures is advantageous. Special devices have been developed for
these applications. In principle, all these algorithms analyze the time-varying spectral distribution of the EEG individually for each channel. The critical parameters extracted from this spectral analysis
are plotted in terms of a trend curve, thus providing significant data reduction and noise removal.
However, a human interpreter still has to make the final decision regarding the presence of pathological
changes in brain activity.
References
American Electroencephalographic Society, Guideline thirteen: guidelines for standard electrode position
nomenclature, J. Clin. Neurophysiol., 11, 111–113, 1994.
Cooper, R., Osselton, J.W., and Shaw, J.C., EEG Technology , 3rd ed., Butterworth-Heinemann, Oxford,
1981.
International Electrotechnical Commission, Geneva, Switzerland, IEC 601 standard “Medical electrical
equipment,” Part 2-26: Particular requirements for the safety of electroencephalographs (IEC 601-
2-26), 1994.Jasper, H.H., The ten-twenty system of the International Federation, Electroencephalogr.
Clin. Neurophysiol ., 10, 371–375, 1958.
Niedermeyer, E. and Lopes da Silva, F., Electroencephalography: Basic Principles, Clinical Applications, and
Related Fields, 4 th ed., Williams & Wilkins, Baltimore, 1998.
Nuwer, M.R., Comi, G., Emerson, R., Fuglsang-Frederiksen, A., Guerit, J.M., Hinrichs, H., Ikeda, A.,
Luccas, F.J.C., and Rappelsberger, P., IFCN standards for digital recording of clinical EEG, Electro-
encephalogr. Clin. Neurophysiol ., 106, 259–261, 1998.
Osselton, J.W., The influence of bipolar and unipolar connection on the net gain and discrimination of
EEG amplifiers, Am. J. EEG Technol ., 5, 53, 1965.
Perrin, F., Pernier, J., Bertrand, O., Giard, M.H., and Echallier, J. F., Mapping of scalp potentials by surface
spline interpolation, Electroencephalogr. Clin. Neurophysiol., 66, 75–81, 1987.
Pichlmayr, I. and Lips, U., EEG monitoring in anesthesiology and intensive care, Neuropsychobiology ,
10(4), 239–248, 1983.
Rechtschaffen, A. and Kales, A., A manual of standardized terminology, techniques and scoring system for
sleep stages of human subjects, Natl. Inst. Neurol. Dis. Blind (NIH Publ. 204), Bethesda, MD, 1968.
Additional Recommendations, Standards, and Further Reading
American Electroencephalographic Society, Guidelines for writing EEG reports, J. Clin. Neurophysiol., 1,
219–222, 1984.
Chatrian, G.E., Bergamasco, B., Bricolo, A., Frost, J., and Prior, P., IFCN recommended standards for
electrophysiologic monitoring in comatose and other unresponsive states, Electroencephalogr. Clin.
Neurophysiol ., 99, 103–126, 1966.
© 2004 by CRC Press LLC
7/21/2019 Electroencephalography
http://slidepdf.com/reader/full/electroencephalography-56dd8d2fbfe6d 21/21
Ebner, A., Sciarretta, C.M., Epstein, C.M., and Nuwer, M., EEG instrumentation, in Recommendations
for the Practice of Clinical Neurophysiology: Guidelines of the International Federation of Clinical
Neurophysiology , Deuschl, G. and Eisen, A., Eds., Suppl. 52 to Electroencephalogr. Clin. Neurophys-
iol ., Elsevier, Amsterdam, 1999.
Noachtar, S., Binnie, C., Ebersole, J., Maugière, F., Sakamoto, A., and Westmoreland, B., A glossary of
terms most commonly used by clinical electroencephalographers and proposal for the report form
for the EEG findings, in Recommendations for the Practice of Clinical Neurophysiology: Guidelines
of the International Federation of Clinical Neurophysiology , Deuschl, G. and Eisen, A., Eds., Suppl.
52 to Electroencephalogr. Clin. Neurophysiol ., Elsevier, Amsterdam, 1999.
Nunez, P.L., Electrical Fields of the Brain, Oxford University Press, New York, 1981.
Nuwer, M.R., Quantitative EEG. I. Techniques and problems of frequency analysis and topographic
mapping, J. Clin. Neurophysiol ., 5, 1–43, 1988.
Nuwer, M.R., The development of EEG brain mapping, J. Clin. Neurophysiol ., 7, 459–471, 1990.
A comprehensive set of further recommendations is available via the Internet from the International
Federation of Clinical Neurophysiology (IFCN)-home page at http://www.ifcn.info/.